database/mariadb
1
0
Fork 0
mirror of https://github.com/MariaDB/server.git synced 2025-07-27 18:02:13 +03:00
Code Activity

Introduce analyze_sample_percentage variable

Browse Source 
The variable controls the amount of sampling analyze table performs.

If ANALYZE table with histogram collection is too slow, one can reduce the
time taken by setting analyze_sample_percentage to a lower value of the
total number of rows.
Setting it to 0 will use a formula to compute how many rows to sample:

The number of rows collected is capped to a minimum of 50000 and
increases logarithmically with a coffecient of 4096. The coffecient is
chosen so that we expect an error of less than 3% in our estimations
according to the paper:
"Random Sampling for Histogram Construction: How much is enough?”
– Surajit Chaudhuri, Rajeev Motwani, Vivek Narasayya, ACM SIGMOD, 1998.

The drawback of sampling is that avg_frequency number is computed
imprecisely and will yeild a smaller number than the real one.
This commit is contained in:
Vicențiu Ciorbaru
2019-02-15 01:23:00 +02:00
parent 47f15ea73c
commit f0773b7842
8 changed files with 265 additions and 10 deletions
Download Patch File Download Diff File Expand all files Collapse all files

104
mysql-test/main/statistics.result
View File

@ -1760,3 +1760,107 @@ DROP TABLE t1;
# End of 10.2 tests
#
set histogram_size=@save_hist_size, histogram_type=@save_hist_type;
#
# Start of 10.4 tests
#
#
# Test analyze_sample_percentage system variable.
#
set @save_use_stat_tables=@@use_stat_tables;
set @save_analyze_sample_percentage=@@analyze_sample_percentage;
set session rand_seed1=42;
set session rand_seed2=62;
set use_stat_tables=PREFERABLY;
set histogram_size=10;
CREATE TABLE t1 (id int);
INSERT INTO t1 (id) VALUES (1), (1), (1), (1), (1), (1), (1);
INSERT INTO t1 (id) SELECT id FROM t1;
INSERT INTO t1 SELECT id+1 FROM t1;
INSERT INTO t1 SELECT id+2 FROM t1;
INSERT INTO t1 SELECT id+4 FROM t1;
INSERT INTO t1 SELECT id+8 FROM t1;
INSERT INTO t1 SELECT id+16 FROM t1;
INSERT INTO t1 SELECT id+32 FROM t1;
INSERT INTO t1 SELECT id+64 FROM t1;
INSERT INTO t1 SELECT id+128 FROM t1;
INSERT INTO t1 SELECT id+256 FROM t1;
INSERT INTO t1 SELECT id+512 FROM t1;
INSERT INTO t1 SELECT id+1024 FROM t1;
INSERT INTO t1 SELECT id+2048 FROM t1;
INSERT INTO t1 SELECT id+4096 FROM t1;
INSERT INTO t1 SELECT id+9192 FROM t1;
#
# This query will should show a full table scan analysis.
#
ANALYZE TABLE t1;
Table Op Msg_type Msg_text
test.t1 analyze status Engine-independent statistics collected
test.t1 analyze status OK
select table_name, column_name, min_value, max_value, nulls_ratio, avg_length, avg_frequency,
DECODE_HISTOGRAM(hist_type, histogram)
from mysql.column_stats;
table_name column_name min_value max_value nulls_ratio avg_length avg_frequency DECODE_HISTOGRAM(hist_type, histogram)
t1 id 1 17384 0.0000 4.0000 14.0000 0.082,0.086,0.086,0.086,0.086,0.141,0.086,0.086,0.086,0.086,0.086
set analyze_sample_percentage=0.1;
#
# This query will show an innacurate avg_frequency value.
#
ANALYZE TABLE t1;
Table Op Msg_type Msg_text
test.t1 analyze status Engine-independent statistics collected
test.t1 analyze status Table is already up to date
select table_name, column_name, min_value, max_value, nulls_ratio, avg_length, avg_frequency,
DECODE_HISTOGRAM(hist_type, histogram)
from mysql.column_stats;
table_name column_name min_value max_value nulls_ratio avg_length avg_frequency DECODE_HISTOGRAM(hist_type, histogram)
t1 id 111 17026 0.0000 4.0000 1.0047 0.039,0.098,0.055,0.118,0.078,0.157,0.082,0.118,0.094,0.063,0.098
#
# This query will show a better avg_frequency value.
#
set analyze_sample_percentage=25;
ANALYZE TABLE t1;
Table Op Msg_type Msg_text
test.t1 analyze status Engine-independent statistics collected
test.t1 analyze status Table is already up to date
select table_name, column_name, min_value, max_value, nulls_ratio, avg_length, avg_frequency,
DECODE_HISTOGRAM(hist_type, histogram)
from mysql.column_stats;
table_name column_name min_value max_value nulls_ratio avg_length avg_frequency DECODE_HISTOGRAM(hist_type, histogram)
t1 id 1 17384 0.0000 4.0000 3.5736 0.082,0.086,0.086,0.082,0.086,0.145,0.086,0.086,0.082,0.086,0.090
set analyze_sample_percentage=0;
#
# Test self adjusting sampling level.
#
ANALYZE TABLE t1;
Table Op Msg_type Msg_text
test.t1 analyze status Engine-independent statistics collected
test.t1 analyze status Table is already up to date
select table_name, column_name, min_value, max_value, nulls_ratio, avg_length, avg_frequency,
DECODE_HISTOGRAM(hist_type, histogram)
from mysql.column_stats;
table_name column_name min_value max_value nulls_ratio avg_length avg_frequency DECODE_HISTOGRAM(hist_type, histogram)
t1 id 1 17384 0.0000 4.0000 7.4523 0.082,0.090,0.086,0.082,0.086,0.145,0.086,0.082,0.086,0.086,0.086
#
# Test record estimation is working properly.
#
select count(*) from t1;
count(*)
229376
explain select * from t1;
id select_type table type possible_keys key key_len ref rows Extra
1 SIMPLE t1 ALL NULL NULL NULL NULL 229060
set analyze_sample_percentage=100;
ANALYZE TABLE t1;
Table Op Msg_type Msg_text
test.t1 analyze status Engine-independent statistics collected
test.t1 analyze status Table is already up to date
select table_name, column_name, min_value, max_value, nulls_ratio, avg_length, avg_frequency,
DECODE_HISTOGRAM(hist_type, histogram)
from mysql.column_stats;
table_name column_name min_value max_value nulls_ratio avg_length avg_frequency DECODE_HISTOGRAM(hist_type, histogram)
t1 id 1 17384 0.0000 4.0000 14.0000 0.082,0.086,0.086,0.086,0.086,0.141,0.086,0.086,0.086,0.086,0.086
explain select * from t1;
id select_type table type possible_keys key key_len ref rows Extra
1 SIMPLE t1 ALL NULL NULL NULL NULL 229376
set use_stat_tables=@save_use_stat_tables;
drop table t1;